Re: [Regression 2.6.35-rc1?] Sysrq works too well (no need of alt)

Re: [Regression 2.6.35-rc1?] Sysrq works too well (no need of alt)

[Rafael J. Wysocki]

On Tuesday, June 08, 2010, Éric Piel wrote:
> Hello,
> I haven't investigated much yet, but I have the feeling that since
> 2.6.35-rc1 my "Print Screen/SysRq" key works only as SysRq: if I press
> it (normally assigned to take screenshot in gnome), nothing happens (no
> input event received in userspace) and no key on the keyboard works
> afterwards, until I press Alt. Actually the keys work, but behave as if
> the sysrq key was kept pressed (can be seen in dmesg, or by pressing "b").
> 
> Looking at the log, a potential culprit is commit
> 97f5f0cd8cd0a05449cbb77d1e6f02e026875802 (Input: implement SysRq as a
> separate input handler). Probably the logic of "have to press all the
> keys at the same time" changed to "have to press the keys one after each
> other". So pressing alt and later on pressing PrintScreen leads to a SysRq.
> 
> Does anybody else see this behaviour? Any suggestion on how to solve
> this bug? I'll try reverting the commit and report if it fixes the
> problem (the git revert fails so I've got to fix the conflict manually).

Is this still a problem with 2.6.35-rc3?

Rafael

Re: [Regression 2.6.35-rc1?] Sysrq works too well (no need of alt)

[Éric Piel]

Op 19-06-10 23:16, Rafael J. Wysocki schreef:
> On Tuesday, June 08, 2010, Éric Piel wrote:
>> Hello,
>> I haven't investigated much yet, but I have the feeling that since
>> 2.6.35-rc1 my "Print Screen/SysRq" key works only as SysRq: if I press
>> it (normally assigned to take screenshot in gnome), nothing happens (no
>> input event received in userspace) and no key on the keyboard works
>> afterwards, until I press Alt. Actually the keys work, but behave as if
>> the sysrq key was kept pressed (can be seen in dmesg, or by pressing "b").
>>
>> Looking at the log, a potential culprit is commit
>> 97f5f0cd8cd0a05449cbb77d1e6f02e026875802 (Input: implement SysRq as a
>> separate input handler). Probably the logic of "have to press all the
>> keys at the same time" changed to "have to press the keys one after each
>> other". So pressing alt and later on pressing PrintScreen leads to a SysRq.
>>
>> Does anybody else see this behaviour? Any suggestion on how to solve
>> this bug? I'll try reverting the commit and report if it fixes the
>> problem (the git revert fails so I've got to fix the conflict manually).
> 
> Is this still a problem with 2.6.35-rc3?
The bug is still in 2.6.35-rc3. The patch that fixes this regression is
in the "for-linus" branch of the input tree, but I think Dmitry hasn't
yet sent a pull request to Linus.

Eric

Re: [Regression 2.6.35-rc1?] Sysrq works too well (no need of alt)

[Dmitry Torokhov]

On Sun, Jun 20, 2010 at 01:15:50AM +0200, Éric Piel wrote:
> Op 19-06-10 23:16, Rafael J. Wysocki schreef:
> > On Tuesday, June 08, 2010, Éric Piel wrote:
> >> Hello,
> >> I haven't investigated much yet, but I have the feeling that since
> >> 2.6.35-rc1 my "Print Screen/SysRq" key works only as SysRq: if I press
> >> it (normally assigned to take screenshot in gnome), nothing happens (no
> >> input event received in userspace) and no key on the keyboard works
> >> afterwards, until I press Alt. Actually the keys work, but behave as if
> >> the sysrq key was kept pressed (can be seen in dmesg, or by pressing "b").
> >>
> >> Looking at the log, a potential culprit is commit
> >> 97f5f0cd8cd0a05449cbb77d1e6f02e026875802 (Input: implement SysRq as a
> >> separate input handler). Probably the logic of "have to press all the
> >> keys at the same time" changed to "have to press the keys one after each
> >> other". So pressing alt and later on pressing PrintScreen leads to a SysRq.
> >>
> >> Does anybody else see this behaviour? Any suggestion on how to solve
> >> this bug? I'll try reverting the commit and report if it fixes the
> >> problem (the git revert fails so I've got to fix the conflict manually).
> > 
> > Is this still a problem with 2.6.35-rc3?
> The bug is still in 2.6.35-rc3. The patch that fixes this regression is
> in the "for-linus" branch of the input tree, but I think Dmitry hasn't
> yet sent a pull request to Linus.
> 

The fix should be in mainline now.

-- 
Dmitry